A columnar db (https://datalance.net/blog/columnar-database/) might be better suited for analytical workloads, where queries typically retrieve a few columns from millions of rows. Row-based systems are often better suited to transactional applications, where every row is often accessed, added or modified. The right choice is determined based on accessing the data. For instance, workloads that require frequent transactions are likely to be better suited to row storage, but workloads that are reporting and aggregating may be much better suited to column storage. I would compare the setups of the queries, amount of data and how often it gets updated before making a decision.